Modeling and Prediction
-
Upload
mohamedhazem -
Category
Documents
-
view
218 -
download
0
Transcript of Modeling and Prediction
-
8/3/2019 Modeling and Prediction
1/9
Proceedings of the Fortieth Turbomachinery Symposium
September 12-15, 2011, Houston, Texas
MODELING AND PREDICTION OF SIDESTREAM INLET PRESSURE FOR MULTISTAGE CENTRIFUGAL
COMPRESSORS
Jay Koch
Principal Engineer
Dresser-Rand
Olean, NY, USA
Jim Sorokes
Principal Engineer
Dresser-Rand
Olean, NY, USA
Moulay Belhassan
Principal Engineer
Dresser-Rand
Olean, NY, USA
.
ABSTRACT
It is common for some compressors in certain applications
to have one or more incoming sidestreams that introduce flow
other than at the main inlet to mix with the core flow. In most
cases, the pressure levels at these sidestreams must be
accurately predicted to meet contractual performance
guarantees. The focus of this paper is the prediction of
sidestream flange pressure when the return channel outlet
conditions are provided. A model to predict the impact of local
curvature in the mixing section is presented and compared with
both Computational Fluid Dynamics work and measured test
data.
INTRODUCTION
Despite recent advances in analytical tools, developers and
users of state-of-the-art centrifugal compressor equipment
continue to rely heavily on testing to ultimately confirm the
performance of new components or stages. This is especially
true in heavy hydrocarbon applications such as compressors
used in liquefied natural gas (LNG), ethylene, or gas-to-liquid
facilities. Heavy hydrocarbon gases have very low gas sonic
velocities that produce high Mach numbers in the aerodynamic
flowpath. By their nature, such high Mach number, high flow
coefficient stages have very narrow flow maps characterized by
limited choke and surge margin. In addition, many of theseapplications require the machines with sidestreams (or
sideloads) to accommodate the flows entering and/or exiting
the compressor as determined by the particular process for
which they are intended. The sidestreams and associated
mixing further complicate the performance prediction process
because the pressure, temperature and flow conditions at each
one of these sidestreams as well as at the exit of the machine
must be met within stringent tolerances to optimize the amount
of end product delivered by the process. Therefore, it is
imperative that the compressor OEM and end user have a firm
grasp of the performance characteristics of any impellers
diffusers, return channels, and/or other flow path components
used in such equipment.
The high cost associated with delays in the projec
schedule increase the attractiveness of design and testing
methods that ensure these machines meet the contractua
operating requirements the first time; without the need for
repeated modifications and test iterations to correc
performance shortfalls. Given the size, construction style, and
in-house test piping arrangements for many large centrifugals
one modification / re-test iteration could take several weeks
Repeated iterations on the OEMs test facility could delay the
start-up of a new plant by months, leading to lost production /
revenue for the end user as well as lost revenue and reputationfor the compressor OEM. Consequently, OEMs spend
continually work to enhance their performance prediction
methods. This paper details work done toward that end.
This paper is the latest in a continuing series of publications
describing work completed in association with a novel test
vehicle that was designed to replicate the performance of a full-
scale, large frame-size, multi-stage centrifugal compressor for
high flow coefficient, high Mach number stages. The test
vehicle, described by Sorokes et al (2009), is equipped with a
vast array of internal instrumentation that make it possible to
evaluate the aero/thermodynamic and mechanical behavior of a
compressor. Sufficient instrumentation is installed to allow
assessment of the entire compressor as well as individualcomponents or combinations of components. Of particular
interest to this study, the test rig included instrumentation at
critical locations within the sidestream components to permit an
assessment of the losses through each sidestream element.
BACKGROUND
As part of the compressor selection and design process, the
OEM must be able to predict the performance of the overall
137
-
8/3/2019 Modeling and Prediction
2/9
unit from inlet to discharge flange. For a simple, so-called
straight through compressor, this means predicting the
performance for each stage (inlet guide / impeller / diffuser /
return channel or volute) in series from the inlet to the
discharge of the machine. However, for units with one or more
incoming sidestreams, there is the additional task of accounting
for the impact of the sidestream and thereby predicting the
conditions at the intermediate flange; i.e., static and total
pressure.
Typically, OEMs supply compressor performance curves
that reflect flange-to-flange performance because that is what
the process engineer needs to ensure proper operation of their
system. Such data is easier to monitor when a compressor is
installed in the field. However, flange-to-flange data, if not
interpreted properly can lead to misleading conclusions as to
the relative performance of individual sections of a
compression system. For example, if the sidestream losses
from the flange to mixing section are attributed to the upstream
section, it will cause the upstream section to appear low in
performance while the downstream section will show
erroneously high performance levels; i.e., efficiency
significantly higher than 90%). The need for a more refined
method for allocating sidestream losses as well as the need to
eliminate other uncertainties regarding the sidestream lossmechanisms has led to additional research on sidestream
performance modeling.
Compressor OEMs have completed significant work in
recent years to improve their understanding of the flow in
sidestreams and sidestream mixing section as well as the impact
of the mixed stream on the downstream impeller. In most
cases, the portion of the sidestream from the flange connection
to the mixing section (i.e., the location where the incoming
sidestream flow mixes with the core flow) is very similar to a
radial compressor inlet. There has been significant work
reported in the literature on radial inlets. For example, Flathers
et al (1994) compared several geometric variations using
Computational Fluid Dynamics (CFD) and measured testresults. Koch et al (1995), Kim and Koch (2004), Michelass
and Giachi (1997) subsequently reported similar work on
different geometric variations. These studies all led to
increased understanding of the flow structure and improved
prediction methods for inlet losses and the details of the flow
field exiting these components. Proper use of the CFD tools
also led to more efficient / effective inlet configurations that
provided a more circumferentially uniform flow to the
downstream impeller. The studies also aided in the one-
dimensional prediction of pressure losses.
Additional studies, which included the upstream return
channel and the complete sidestream geometry, have been
completed to accurately model the flow where the core flow
and sidestream flow merge. While this location is typically
called the mixing section as noted above, the flows do not
completely mix before entering the impeller. The works of
Sorokes et al (2000 and 2006) and the prior referenced work on
radial inlets have led to a greater comprehension of the
complex flows at the sidestream mixing location. However,
these works did not address the flow physics that determine the
static pressure in the mixing section, which, in turn, sets the
flange pressure level. The study by Hardin (2002) describes a
one-dimensional method that determines how the flow at the
mixing location and, therefore, the local static pressure is
impacted by the local flow curvature.
This paper presents predictions for two geometric
configurations that violate the assumptions made in previous
studies, resulting in a need to modify the relevant equations
Additionally, the prior work did not present the methods used to
predict the total pressure downstream of the mixing section
This paper will present various methods to predict the
downstream total pressure and compare the analytical results to
measured test data on the two geometries.
INVESTIGATIVE STUDY
As part of a recent OEM development program, a study on
several sidestream configurations was conducted. The projec
required compact sidestream mixing sections; i.e., mixing
sections that take less axial length; in a compressor with high
flow coefficient ( > 0.10) and high machine Mach number
(i.e., U2/A0 > 1.1). The sidestreams were of different sizes a
the incoming flows represented different percentages of the
flow from the upstream section. The gas density in the
sidestreams also varied due to the increasing pressure in the
machine. Finally, the machine was tested using a high mole
weight refrigerant to replicate the high gas densities and Machnumbers. Representative geometry for a typical configuration
is shown in Figures 1a and 1b.
Figure1a. Geometry for configuration 1
Figure1b. Geometry for configuration 1 section A- A
138
-
8/3/2019 Modeling and Prediction
3/9
During the detailed design process, each proposed stage
geometry was modeled using a commercially available CFD
code, ANSYS CFX-10. As part of this study, the sidestream
inlet geometry was optimized within the available axial space
provided. The CFD domain for each sidestream included the
upstream diffuser, return channel, sidestream nozzle, sidestream
plenum, mixing section, and downstream impeller. The full
360-degree model included all passages to evaluate any
circumferential flow non-uniformity at the exit of the
sidestream. The upstream impeller was not included to reduce
the size of the computational domain, thereby reducing solution
time. This simplification also allowed the return channel exit
conditions to be varied for different flow conditions; i.e., to
replicate the flow conditions from maximum (choke) to
minimum stable flow (stall / surge). Because the objective of
the study was to predict the sidestream flange pressure, when
provided return channel exit conditions, it was felt this
simplification would have a minimal impact on the flow
physics of interest.
The geometry was modeled using a combination of
hexahedral and tetrahedral elements. All axisymmetric
components, such as the diffuser and return channel, were
modeled using hexahedral elements and the non-uniform
components, such as the sidestream nozzle and plenum section,were modeled using tetrahedral elements because such
components are more easily modeled using tets. The grids
were refined for each component and grid element type as it is
recognized that higher grid densities are required for tetrahedral
as opposed to hexahedral elements. Each component had a grid
size of approximately 2 million elements and the total grid size
was approximately 6 million elements. For the sidestream
inlet, the grid density used is similar to those reported by the
current authors in Kim and Koch (2004).
The boundary conditions of the model were based on the
one-dimensional predictions used to select and size the
hardware. The prescribed flow angle, total pressure and totaltemperature were specified at the diffuser entrance. Total
temperature and mass flow were specified at the sidestream
entrance and total mass flow was specified at the exit of the
computational domain. The inlet pressure at the sidestream
entrance was the objective function of the analysis.
A second order discretization scheme was used for the
simulation and standard k- model was adopted as a turbulence
model combined with a scalable wall function approach that
eliminates the necessity of discretely resolving the large
gradients in the thin, near-wall region. The convergence
criterion was set to the maximum residual of 10-4
for u, v, w,
and p.
For the two sidestream configurations addressed in thispaper, the internal elements, return channel vanes, sidestream
vanes and inlet guide vanes, were fixed during the optimization.
Only the meridional passage width was varied.
In the first configuration (designated Configuration 1), the
core flow (i.e., flow exiting the upstream return channel) and
incoming sidestreams flows were nearly equal. This geometry
is shown in Figures 1a and 1b. The second configuration
(Configuration 2) was for a higher pressure and hence higher
density stage and had a core flow that was much greater than
the sidestream flow. The details of this mixing section
geometry can be seen in Figure 2.
Figure 2. Mixing section geometry for configuration 2
The design values of area ratio (return channel exit area to
sidestream exit area), density, and return channel exit Mach
number are provided in Table 1.
Table 1. Sidestream Design Parameters
The results of the CFD study are shown in Figures 3-6
The Mach number distribution in the sidestream nozzle for
configuration 1 is shown in Figure 3. This configuration show
the vane rows in the sidestream are oriented with the flow to
provide a circumferentially uniform velocity profile at the exit
of the sidestream. A meridional view providing the Mach
number distribution in the mixing section is shown in Figure 4This plot shows a variation in the Mach number across the
sidestream exit. The variation in Mach number is also eviden
at the exit of the return channel as the flow turns to enter the
downstream impeller.
Density Area Ratio Mach
lb/ft3
Main/Sidestream Number
1 0.30 0.76 0.34
2 1.00 4.96 0.26
Configuration
139
-
8/3/2019 Modeling and Prediction
4/9
Figure 3. Mach number in sidestream configuration 1
Figure 4. Mach number at sidestream mixing location -
configuration 1
Figure 5. Static pressure at the sidestream mixing location -
configuration 1
Figure 6. Static pressure at the sidestream mixing location
configuration 2
Contours of the static pressure at the same location are
shown in Figure 5. It is evident from this figure that the static
pressure varies from the shroud wall at the exit of the
sidestream to the hub wall downstream of the return channel
Because of the lower static pressure at the sidestream exit, the
total pressure at this location is lower even though the velocity
upstream of the curvature is roughly matched for both streams.
This phenomenon of reduced pressure at the sidestream
exit was predicted by Hardin (2002). Based on the
relationships shown in that work, this variation in pressure can
be expected to increase with dynamic pressure, 22V , and
with larger variations in curvature between the two streams.
A plot of static pressure for configuration 2 is shown in
Figure 6. This configuration shows similar trends to
configuration 1 with significant variation in the static pressure
from the exit of the return channel to the exit of the sidestream.
In reviewing these configurations it is evident that the
specific formulation developed in the prior work does not apply
to this geometry. A major assumption put forth in the prior
work was that there was no variation in the static pressure at the
outlet of the return channel passage. Thus the effect o
curvature was only applied across the sidestream exit. For this
geometry, the formulation must be modified to account for the
variation across both the return channel exit and the sidestream
exit. While simple in nature, this can have a significant impac
on the predicted static pressure at the sidestream exit.
Before developing a method to account for the changes in
curvature, an optimization study was conducted on the mixing
section of configuration 2. To reduce the execution time and
allow for more design iterations, a sub-model was created that
included a portion of the return channel vane exit and a portion
of the exit of the plenum section of the sidestream just upstream
of the mixing section. The upstream diffuser and return
channel as well as the sidestream plenum were removed from
the sub-model. This also allowed the geometry to be modeled
as a pie slice section as opposed to a full 360 model
Comparison of sub-model and full model solutions revealedthat the simplified model captured the important flow features
in the mixing section. Therefore, the simplified model was
used for the optimization study for the mixing section.
PREDICTION METHODOLOGY
To accurately predict the sidestream exit conditions, proper
boundary conditions for the problem must be established. The
exit conditions at the return channel are generally known from
one-dimensional predictions or from test measurements
Therefore, for this discussion it was assumed the discharge tota
pressure was known at the exit of the return channel preceding
the mixing section. Additionally, it is assumed the area, flowrate, temperature and gas properties were known at the return
channel exit and at the sidestream exit.
Given the referenced information, the return channel exi
mean velocity and static pressure can be calculated. Based on
basic flow principles, it was assumed that the static pressure
was equal between both flow streams at the mixing location
The only location where this is true is at the shroud wall at the
exit of the return channel and the hub wall at the exit of the
sidestream. This assumption is supported by the previou
computational study completed by the OEM and by other
140
-
8/3/2019 Modeling and Prediction
5/9
researchers such as Hardin (2002). Unfortunately, the pressure
at this location is not as desired because of the local curvatures
in the mixing section. The desired result can be obtained via a
series of equations that allow the average static pressure at the
exit of the sidestream to be calculated from the average static
pressure at the exit of the return channel.
According to basic aerodynamic principles, the
relationship of pressure versus radius of curvature can be
expressed by Equation (1).
R
V
dR
dP2
= (1)
For the geometry shown in Figure 7, the average static
pressure at the exit of the return channel can be related to the
average pressure at the exit of the sidestream by the variation in
curvature using Equation 1. The average static pressure at the
sidestream exit is defined by Equation (2).
( )
t
sst
tss
c
cc
ss
R
RRVPP
Re
Re
Re
2
+=
(2)
Where:
sssP = Static pressure at sidestream exit
tsP
Re
= Static pressure at return channel exit
ssCR = Radius sidestream of curvature (see fig. 7)
tCR
Re
= Radius of return channel curvature (see fig. 7)
= Gas density
V = Velocity at the sidestream exit
Figure 7. Geometry for pressure prediction
Once the static pressure at the sidestream exit is
determined, the velocity and total pressure can be calculated.The flange total pressure can be calculated based on the
predicted losses in the nozzle plenum region, once the
sidestream exit total pressure is known (Equations (3) and (4)).
The losses in this region of the sidestream have been shown to
be consistent and predictable by 1-D models. The 1-D model
for a given geometry can be based on CFD prediction or based
on test measurements.
2
2V
PPssss sT
+= (3)
ss
FLG
TTLC
VPP
ssFLG 2
2
+= (4)
Where:
sssP = Static pressure at sidestream exit
ssTP = Total pressure at sidestream exit
FLGTP = Total pressure at flange
= Gas density
V = Velocity at sidestream exit
VFLG = Velocity at flange
LCss = Loss coefficient
The appropriate method for the prediction of the tota
pressure downstream of the mixing section at the inlet of the
impeller has not been questioned in the literature. The
appendix of the test standard from ASME (1997), defines a
momentum balance method, but this method is approximateAttempts to use this method have not been very successful for
these authors or others in the open literature. The impact of the
gradient in total pressure on the downstream impeller
performance varies greatly. Some CFD studies indicate a
significant impact on performance while other studies show a
minor impact. For 1-D performance prediction purposes a
single representative value of the impeller inlet pressure is
desired. This representative pressure should provide a
consistent reference between predicted performance and
measured test results. It is also desired that the selected value
provides a consistent reference for comparison between
performance prediction and measured test results.
When CFD results are evaluated, a mass-averaged totapressure at a selected plane is the common method for
evaluation. Evaluations of test results often use an arithmetic
average or area average based on total pressure measured by a
single pressure probe or multi-hole pressure rake. The prio
work by Hardin (2002) suggests mass-averaging the tota
pressure at the exit of each stream. To determine which method
provides a more consistent result, a set of evaluation criteria
was developed for use on future development tests. The tes
results will be evaluated using the three methods listed above to
determine which method provides a more consistent result and
therefore used for future comparisons between test results and
analytical predictions.
TEST DATA AND DISCUSSION
After completion of the design study, the OEM constructed
a full-scale development test rig; again, see Sorokes et al
(2009). As noted, the flow path of the test rig was heavily
instrumented to allow the measurement of the aero-
thermodynamic performance of the components of each stage
including the sidestream elements. The interna
instrumentation included total pressure probes, dynamic
141
-
8/3/2019 Modeling and Prediction
6/9
pressure probes, total temperature probes, and five-hole probes,
which measure both pressure and flow direction. A schematic
of the internal instrumentation layout used for each compressor
stage is given in Figure 8.
Figure 8. Test rig internal instrumentation layout
For each sidestream, the total pressure and temperature
were measured at the exit of the return channel and at the
sidestream exit. Additionally, the total pressure and
temperature distributions were measured near the exit of the
mixing section; i.e. just upstream of the impeller; via pressure
and temperature rakes. Therefore, it was possible to assess the
hub-to-shroud variation (or stratification) in pressure and
temperature due to the core / sidestream flow mixing.
In addition to the internal instrumentation, the total
pressure, total temperature and mass flow were measured at the
flange locations using standard combination pressure /
temperature probes.All performance testing was conducted using R-134A
refrigerant to achieve aerodynamic similitude at the design
conditions. The test measurements for each sidestream were
taken in accordance with the ASME PTC-10 test code from
ASME (1997).
To evaluate the revised prediction scheme, the predicted
sidestream exit total pressure was calculated at the test
conditions and compared with the measured total pressure on
test. The prediction was based purely on the known geometry
and test conditions.
For each sidestream configuration, data is presented for
different flow rates at a fixed impeller tip Mach number of
1.10. The flow ratio between the sidestream and the core flow
is maintained for all operating points using the OEM stringent
guidelines; i.e., more restrictive than the standard ASME Power
Test Code (see Kolata and Colby (1990)). The temperatures for
core flow and incoming sidestream streams were kept equal to
avoid heat transfer that would add uncertainty to the efficiency
calculation.
The results of the prediction for configuration 1 are shown
in Figure 9.
0.20 0.21 0.22 0.23 0.24 0.25 0.26 0.27 0.28 0.29 0.30
Return Channel Exit Mach Number
-10
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
%ErrorinSidestreamExitTo
talPressure(Predicted/Measured)
Prediction with Curvature Correction
Prediction without Curvature Correction
Figure 9. Error between test and measurement with and withou
curvature correction configuration 1
The percent error in the sidestream exit predicted tota
pressure versus Mach number at the outlet of the return channe
is shown. For configuration 1 the prediction model is within
1.5% of the test results. As a means of determining the
suitability of this calculation, the percent error in sidestream
total pressure was also calculated without the curvature
correction. This comparison showed the sidestream exit tota
pressure was always over-predicted without the curvature
correction. For this application, failure to account for the
curvature effects resulted in an error of up to 5.8%. Simila
results for configuration 2 are shown in Figure 10.
0.18 0.19 0.20 0.21 0.22 0.23 0.24 0.25 0.26 0.27 0.28
Return Channel Exit Mach Number
-10
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
1
2
3
4
5
%ErrorinSidestreamExitTotalPressure(Predicted/Measured)
Prediction with Curvature CorrectionPrediction without Curvature Correction
Figure 10. Error between test and measurement with and
without curvature correction configuration 2
142
-
8/3/2019 Modeling and Prediction
7/9
For configuration 2, the prediction model accounting for
curvature is also within 1.5% of the test results. The error
without the curvature correction is even greater than on
configuration 1, with a maximum difference of more than 6.1%.
Because there was considerable interest in the performance
of the sidestream and downstream impeller with off-design
flow functions, a set of conditions was defined to investigate
such operating conditions. These additional data provided
insight into prediction of the impeller inlet total pressure
downstream of a sidestream for off-design conditions.
During these tests the unit was intentionally tested at core
flow to sidestream flow ratios that were at the limits of the
ASME test standard (i.e., 10%), and thus considerably
different than the design values. Based on the curvature
prediction model, the flow function variation causes the
sidestream exit pressure (and thus flange pressure) to vary from
the design conditions. This would be true even if the
downstream impeller inlet conditions remain unchanged.
The change in flange pressure due to the different flow
ratios leads to variations in the calculated section performance
even though the impeller / diffuser / return channel
performance remains essentially unchanged. Recall that it was
possible to determine the performance from impeller exit to
return channel exit because of the additional instrumentationavailable in the test vehicle. The internal performance (based
on impeller inlet to return channel exit) and the flange to flange
performance are shown in Figure 11 for the same operating
conditions. There are small changes in the internal
performance for the various sidestream flow ratios, most likely
due to the impact on varying impeller inlet velocity profile due
to mixing effects at the off-design flow ratios. However,
there are more noticeable differences in the flange to flange
calculation, particularly toward the high capacity end of the
map.
0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10 1.15
Normalized Inlet Flow Coefficient
0.40
0.50
0.60
0.70
0.80
0.90
1.00
1.10
1.20
1.30
1.40
NormalizedHeadCoefficient
Flange-to-Flange -- Design Flow Ratio
Flange-to-Flange -- SS Flow > Design Flow Ratio
Internal -- Design Flow Ratio
Internal -- SS Flow > Design Flow Ratio
Figure 11. Flange to Flange and Internal Head Coefficient for
various sidestream flow ratios
As part of the prediction method discussion, data from
these tests were also used to evaluate which method of
determining the downstream impeller inlet condition provides
the best correlation with the test results. To complete thi
comparison, the impeller inlet condition was calculated three
ways. The first method was based on a mass-average of the
total pressures at the outlet of the return channel and the
sidestream exit. The second method used an arithmetic average
of the total pressure rake downstream of the mixing section
The third method was based on an area average of the
downstream rake pressures.
All three methods were compared using polytropic head
coefficient, p, versus inlet flow coefficient, , (see Figure 12).
0.8 0.85 0.9 0.95 1 1.05 1.1 1.15
Normalized Inlet Flow Coefficient
0.40
0.50
0.60
0.70
0.80
0.90
1.00
1.10
1.20
1.30
1.40
NormalizedHeadCoefficient
Mass Avg - Design
Area Avg - Design
Arithmetic Avg - Design
Figure 12. Internal head coefficient for various total pressure
averaging methods at the mixing section
The data was normalized using the design point head and
flow coefficient and trends were reviewed for both design flow
and off-design flow conditions (i.e., from overload to
stall/surge). For each case, the basic impeller inlet similarity
condition remained the same. That is, the compressor was runat the same tip Mach number, and Reynolds number. However
the sidestream flow function was allowed to vary within the
limits of the ASME test code. As can be seen in the curves, the
different averaging methods did result in a change in the
perceived performance. This is not overly surprising because
the different averaging methods will result in a different inle
pressure to the downstream impeller and because the head
coefficient is directly dependent on the inlet pressure, the
resulting variation in head coefficient is to be expected.
Specifically, the change in inlet pressure due to the various
averaging techniques caused a shift in the inlet capacity and
also caused a change in the head coefficient as well as a change
in rise-to-surge for the area average method. The area-averagedmethod predicts a larger capacity and pressure coefficient than
either method. The mass-averaged result falls between the two
other methods. Based on this comparison, each method would
appear to provide a consistent basis for comparison with
predictions. The mass-averaged method has the advantage o
using a measured quantity that has less variation hub to shroud
than the other methods and is less sensitive to the flow ratio of
the core flow to the sidestream flow. Finally, the mass
averaging method also allows for a direct comparison between
143
-
8/3/2019 Modeling and Prediction
8/9
one-dimensional predictions, CFD and test results, as the
required information is typically readily available.
CONCLUSIONS
This paper presents results from a recent sidestream design
study and compares the analytical results to test measurements.
The new method builds on the prior works of Hardin (2002),
Sorokes et al (2000) and Sorokes et al (2006). Based on this
study, a modification to the published methodology has been
proposed to improve the calculation of the inlet pressure for
standard incoming sidestreams. Using the new method, it is
possible to predict the sidestream exit pressure within an
accuracy of 1.5%. This method enhancement will make it
possible to provide more accurate sidestream pressure levels,
thereby improving the process modeling for new or upgraded
facilities.
The comparison of test results with different averaging
methods has shown that each method offers consistent, though
somewhat different result. Based on this study, it is felt that
mass averaging of the core flow pressure and the sidestream
pressure provides the most effective reference for correlation of
the impeller inlet total pressure.
DISCLAIMER
The information contained in this document consists of
factual data, and technical interpretations and opinions which,
while believed to be accurate, are offered solely for
informational purposes. No representation or warranty is made
concerning the accuracy of such data, interpretations and
opinions.
NOMENCLATURE
V = VelocityR = Radius of Curvature
N = Rotational Speed
D = Impeller Outer Diameter
Q = Volume Flow
U = Tip Speed, DN
n = Polytropic Exponent,
2
1
1
2
ln
P
Pln
= Density
= Specific Volume
= Inlet Flow Coefficient,3
2 ND
Q
= Polytropic Head Coefficient,( )
2
1122
U
P-P1-n
n
PT = Total Pressure
PS = Static Pressure
TT = Total Temperature
Ret = Return Channel Exit
SS = Sidestream Exit
Flg = Sidestream Flange
REFERENCES
ASME, PTC 10, 1997, Performance Test Code on Compressors
and Exhausters, @ ASME Press.
Flathers, M., Bache, G., Rainsberger, R., 1994, An
Experimental and Computational Investigation of Flow in a
Radial Inlet of an Industrial Pipeline Centrifugal Compressor
@ ASME 94-GT-134.
Gilarranz, J., 2007, Actuation and Control of a Movable
Geometry System for a Large Frame-Size, Multi-Stage
Centrifugal Compressor Test Rig, @ ASME Paper GT2007-
27592, Submitted for Presentation at the GT2007, ASME Turbo
Expo 2007, Montreal, Canada.
Hardin, James R., 2002, A New Approach to Predicting
Centrifugal Compressor Sideload Pressure, @ IMECE2002-
39592, Proceedings of ASME International Mechanica
Engineering Congress & Expo, 2002, New Orleans, USA.
Kim, Y., Koch, J., 2004, Design and Numerical Investigation of
Advanced Radial Inlet for a Centrifugal Compressor Stage, @IEMCE2004-60538, Proceedings of ASME Internationa
Mechanical Engineering Congress & Expo, 2004, Anaheim
CA, USA.
Koch, J., Chow, P., Hutchinson, B., Elias, S., 1995
Experimental and Computational Study of a Radial Compressor
Inlet, @ ASME 95-GT-82.
Kolata, J. P., and Colby, G. M., 1990, Performance and
Sidestream Testing, @ Proceedings of the 1990 Dresser-Rand
Technology Seminar, Olean, NY.
Michelass, V., Giachi, M., 1997, Experimental and NumericaAnalysis of Compressor Inlet Volutes, @ ASME 97-GT-481.
Sorokes, J., Kopko, J., Koch, J., 2006, Sidestream optimization
for LNG compressor applications, @ European Fluid
Machinery Congress.
Sorokes, J., Nye, D., DOrsi, N., Broberg, R., 2000, Sidestream
Optimization Through the Use of Computational Fluid
Dynamics and Model Testing, @ Proceedings of the Twenty-
ninth Turbomachinery Symposium, Thurbomachinery
Laboratory, Texas A&M University, College Station, Texas, pp
21-30.
Sorokes, J.M., Soulas, T., Koch, J.M., Gilarranz, J. L., 2009
Full- Scale. Aerodynamic And Rotordynamic Testing For Large
Centrifugal Compressors, @ Proceedings of the Thirty Eighth
Turbomachinery Symposium, Thurbomachinery Laboratory
Texas A&M University, pp. 71-80.
ACKNOWLEDGEMENTS
144
-
8/3/2019 Modeling and Prediction
9/9
The authors thank Dresser-Rand Company for funding this
overall project and for allowing the publication of this work.
145